Abstract
Purpose :
Dynamics of lamina cribrosa (LC) responding to intraocular pressure (IOP) changes are thought to contribute to glaucoma pathogenesis. In order to extract meaningful biomarkers, the effect of pressure on the LC must be readily distinguished from the inherent variability due to scanning conditions, image noise, registration, and segmentation. The purpose of this study is to assess the stability of our new 3D LC micro-structure analysis method using repeated 3D OCT scans on a healthy volunteer.
Methods :
Ten 3D OCT images were acquired from a healthy volunteer scanned with a prototype swept source OCT system. A geometric average template image was estimated by minimizing the amount of deformation which map all images into correspondence. The non-linear deformations (Fig A) encode structural differences between scans. Additionally, 45 templates were estimated using unique image pairs (90 images), as only 2 images are typically acquired. Segmentations of LC beams and pores using multiscale vessel enhancement was done in template space and mapped to observations via deformations. Another volunteer was scanned under elevated IOP using an ophthalmodynamometer.
Results :
Deformations from template to each image (Fig B) show low magnitude (mean ~1 voxel) and consistent distribution, indicating that each imaged LC is structurally similar. Templates from 2 scans (Fig C) are nearly indistinguishable from the template estimated with 10 scans, showing stability with limited images. This is quantified in Fig D; the deformation between templates and repeated observations (0.85±0.5 voxels) is an order of magnitude lower than a patient with artificially elevated IOP (p<0.0001 Wilcoxon signed-rank). Segmented LC beams and pores show a max volume difference of 5.8% and 63 of 90 images are within 2.9% (Dice overlap 0.98±0.006).
Conclusions :
The variability of the LC structure due to imaging and deformable modeling is clearly delineated from the effect of increased IOP. Templates estimated from image pairs show measured LC structural differences are minimal (< 1 voxel) and that 2 scans are sufficient.
This is an abstract that was submitted for the 2018 ARVO Annual Meeting, held in Honolulu, Hawaii, April 29 - May 3, 2018.